In collaboration with Kevin Hall, I have been interested in modeling how the body partitions the macronutrients of the diet into lean and fat body tissue. Using empirical cross-sectional data of the relationship between fat and lean mass, we have developed an expression for how substrate utilization adapts to changes of diet, energy expenditure, and body fat so that energy imbalances produce the observed changes of body composition. The theoretical prediction matches experimental data without the use of free parameters. We used the model to compare the total US food supply with the estimated energy expended by the population and showed that the food supply is exceeding the demand. The results predict that the US obesity epidemic is due to a push effect of an excess supply on food leading to a progressive increase in food waste. Recently, we have analyzed global data and found evidence for a push effect in other countries and in particular developed countries. Hall and I have also shown how measurements of body mass can be used to infer food intake. The model also shows that the time scale to approach steady state when diet is changed is very long, on the order of years. This slow response may partially explain why it is so hard to maintain weight loss. People do not receive timely feedback on their diet performance. We have also resolved the long standing paradox of why body weight is maintained precisely even when food intake wildly fluctuates. We showed that the long time constant smooths the intake fluctuations and used stochastic calculus to calculate a formula. In collaboration with Boyd Swinburn and Nancy Butte, we have generalized the model to apply to growing children. In collaboration with Anne Sumner and Vipul Periwal, we developed a time dependent mathematical model of the suppression of lipolysis by insulin. The model can be used to develop a quantitative measure for insulin's effect on serum free fatty acids akin to insulin sensitivity for glucose. We have tested 23 possible models and used Bayesian model comparison to choose the model that best balances fit to data with minimal model complexity. We were now doing experiments to validate the model and applying it to different populations. We have also used the model to compare differences in free fatty acid dynamics between white and black women. In collaboration with Vipul Periwal, we have developed a new Bayesian scheme to estimate insulin secretion. We have recently augmented the model to account for ingested meals. We have also applied the model to adolescents with Jack Yanovski.